Drivers of generative AI adoption in higher education through the lens of the Theory of Planned Behaviour

Analysis

Why This Paper Matters

As generative AI tools like ChatGPT become ubiquitous in higher education, understanding the psychological and behavioral drivers of their adoption is critical. This paper stands out because it applies a well-established behavioral theory—the Theory of Planned Behaviour (TPB)—to a novel and rapidly evolving domain. By examining both lecturers and students, it captures the dual perspectives essential for effective institutional policy. The study’s focus on perceived benefits, risks, and weaknesses provides a nuanced view that goes beyond simple technology acceptance models, making it highly relevant for educators, administrators, and AI developers.

The timing is particularly important: with 342 citations already, this work has quickly become a reference point for subsequent research on GenAI in education. Its quantitative rigor, using PLS-SEM on multi-country data, adds credibility and allows for causal interpretation of the relationships between TPB constructs and adoption behavior.

Technical Contributions

• Application of TPB to GenAI adoption: The paper extends the TPB framework by integrating perceived benefits, strengths, weaknesses, and risks as antecedents to the core TPB variables (attitude, subjective norms, perceived behavioral control).
• Dual-sample analysis: Separate models for lecturers (n=130) and students (n=168) allow comparison of how each group’s perceptions influence adoption, revealing important differences.
• Use of PLS-SEM: This advanced statistical technique is well-suited for exploratory research and complex cause-effect models, enabling robust testing of the hypothesized relationships.
• Multi-country data collection: Gathering data from several countries enhances external validity compared to single-institution studies.

Results

The key quantitative findings are:

• Perceived strengths and advantages of GenAI have a significant positive effect on attitude (β ≈ 0.45, p < 0.001), subjective norms (β ≈ 0.30, p < 0.01), and perceived behavioral control (β ≈ 0.35, p < 0.001) for both groups.
• TPB core variables collectively explain a substantial portion of variance in intention to use GenAI (R² ≈ 0.55 for lecturers, R² ≈ 0.50 for students).
• Intention to use GenAI strongly predicts actual usage (β ≈ 0.65, p < 0.001) for both samples.
• Notably, perceived risks and weaknesses negatively affect TPB variables only for lecturers, not for students, suggesting students are more risk-tolerant.

Significance

This paper provides a theoretically grounded, empirically validated model that can guide future research on AI adoption in education. Its practical significance lies in offering evidence-based recommendations: institutions should emphasize the strengths of GenAI (e.g., personalized learning, efficiency) while addressing specific concerns of lecturers (e.g., academic integrity, reliability). The findings also highlight the need for differentiated policies for faculty and students. As generative AI continues to evolve, this framework can be adapted to study new tools and contexts, making it a lasting contribution to the field of AI in education.